TLDR Tech / Practical
Netflix's Tiered Storage for Time Series Data
Engineers at Netflix have implemented two data tiering strategies to decrease storage expenses for their petabyte-scale temporal event data. The initial strategy involves streaming data from Apache Kafka and Flink into S3, formatted as optimized Parquet files. The second strategy utilizes native reading from Cassandra S3 backups, employing pluggable Memtables and a custom compaction approach.
Read the full story